Mapping vascular network architecture in primate brain using ferumoxytol-weighted laminar MRI
- Laboratory for Brain Connectomics Imaging, RIKEN Center for Biosystems Dynamics Research, Japan
- Department of Radiology, Washington University in St. Louis, United States
- Department of Neuroscience, Washington University in St. Louis, United States
- Siemens Healthcare K.K., Japan
Figures
Figure 1 with 1 supplement
Ferumoxytol-weighted MRI reveals heterogeneous vascularity in the macaque brain.
Representative 3D gradient-echo images (A) before and (B) after the ferumoxytol contrast agent injection. (C) Ferumoxytol-induced change in transverse relaxation rate (ΔR2*) displayed on subcortical gray matter and cortical midthickness surface contour (N = 1). Average (D) pre-ferumoxytol R2* and (E) ΔR2* equivolumetric layers (ELs; N = 4). (F) Histograms in selected brain regions and (G) ELs. Solid lines and shadow indicate mean and standard deviation (N = 4), respectively.
Figure 1—figure supplement 1
Transverse relaxation rate (R2*) measures are biased by the orientation of the static magnetic field (B0).
(A) The angle (θ) between B0 and the normal of the cortex. (B) Representative pre-ferumoxytol R2* (left) and ferumoxytol-induced change in R2* (ΔR2*; right) plotted with respect to the cosine squared of θ from an equivolumetric layer 4a (EL4a) (N = 1). Note that ΔR2* is high when the cortex and B0 are perpendicular and low when they are parallel. (C) B0 orientation bias exhibits laminar depth dependence. Values indicate mean and error bars indicate standard deviation across subjects (N = 4). Interestingly, R2* is positively whereas ΔR2* is negatively correlated with B0 orientation. R2* bias may reflect diamagnetic myelin sheath enwrapping axons that are oriented mainly parallel to the normal of cortex whereas ΔR2* may reflect vessels (e.g., arterioles, capillaries, and venules) with a net orientation perpendicular to the normal of the cortex.
Figure 2 with 3 supplements
Charting large-caliber vessel networks in the cerebral cortex.
(A) Ferumoxytol-weighted MRI reveals a continuous pial vessel network running along the cortical surface. Note that the large vessels branch into smaller pial vessels. (B) Cortical surface mapping of intra-cortical vessels. Vessels were identified using high-frequency gradients (red-yellow colors) and each blue dot indicates the vessel’s central location. Representative equivolumetric layer (EL) 4a is displayed on a 656k surface mesh. (C) Number of penetrating vessels across ELs per hemisphere. Solid lines and shadow show mean and standard deviation across TEs (N = 1). (D) Non-uniformly sampled Lomb–Scargle geodesic-distance periodogram. The vessels exhibit a peak frequency at about 0.6 1/mm reflecting the frequency of large-caliber vessels. (E) Comparison of vessel density in V1 determined using MRI (current study) and ‘ground-truth’ anatomy. In volume space, the density of vessels was estimated using Frangi filter whereas in the surface mesh the density was estimated using local minima. These are compared to the density of penetrating vessels with a diameter of 20–50 μm (Zheng et al., 1991) and the density of feeding arterial and draining veins evaluated using fluorescence microscopy (Weber et al., 2008).
Figure 2—figure supplement 1
Consistency of pial vessel network mapping across subjects.
(A) After ferumoxytol contrast agent injection, pial vessels induce continuous signal-losses in the superficial equivolumetric layer 1b (EL1b). (B) Continuous signal-dropouts are largely absent in deeper cortical layers such as EL4b.
Figure 2—figure supplement 2
Charting vessels in the visual cortex.
(A) Ferumoxytol-weighted gradient-echo image (TE = 14 ms, 0.23 mm isotropic). The yellow arrow highlights a layer with high vascular density, likely corresponding to the primary input layer IVc. The dark gray arrow indicates a penetrating vessel extending to the white matter. Snippet shows the location of the zoomed view. (B) Vessel detection using Frangi filter. The green arrow denotes a pial vessel running along the cortical surface. (C) Vessel detection using low-frequency subtracted signal-intensity surface maps. Blue color signifies the central location of a vessel in a representative equivolumetric layer 3b. Note that not all vessels are labeled in this view as some of them are aligned orthogonal to the slice and their peaks are located in the adjacent imaging slice.
Figure 2—figure supplement 3
Detection of intra-cortical vessels across equivolumetric layers.
Figure 3 with 1 supplement
Exemplar laminar profiles of transverse relaxation rate (R2*) and ferumoxytol-induced change in R2 (ΔR2*) in the macaque cerebral cortex.
(A) Exemplar equivolumetric layer 4b (EL4b) R2* and (B) ΔR2* displayed on cortical flat-map. Note that primary sensory areas (e.g., V1, A1, and area 3) and association areas exhibit high and low ΔR2*, respectively. (C, D) Exemplar laminar profiles from visual cortical areas. Solid lines and shadow show mean and standard deviation across hemispheres, respectively. (E) Laminar ΔR2* profiles relative to the V1. Solid lines and shadow show mean and inter-subject standard deviation. (F) Peak-normalized ΔR2* profile compared with anatomical ground-truth in V1 (Weber et al., 2008). Cytochrome-c oxidase (CO) activity, capillary and large vessel volume fractions were estimated from their Figure 4. Abbreviations: A1: primary auditory cortex; A7: Brodmann area 7; MT: middle temporal area; V1: primary visual cortex; V2: secondary visual cortex; 3: primary somatosensory cortex; 4: primary motor cortex.
Figure 3—figure supplement 1
Ferumoxytol-induced change in R2 (ΔR2*) across equivolumetric layers (ELs).
Data was parcellated using M132 atlas.
Figure 4
Variations in vascular network architecture reveal cortical area boundaries.
(A) Ferumoxytol-induced change in transverse relaxation rate (ΔR2*) displayed at a representative equivolumetric layer 4b (EL4b) (N = 4). Overlaid black lines show exemplary M132 atlas area boundaries. (B) ΔR2* gradients co-align with exemplary areal boundaries. Red arrow indicates an artifact from inferior sagittal sinus. Average (C) mid-thickness weighted T1w/T2w-FLAIR myelin and (D) cortical thickness maps.
Figure 5
Hierarchical organization and principal types of cerebral vasculature.
(A) Average ΔR2* equivolumetric layers (ELs) ascending from pial surface (left) to white matter surface (right) (N = 4; hemispheres = 8). Parcel order was sorted by (B) dendrogram determined using Wards’ method. (C) Similarity matrix as estimated using Euclidean distance. (D) Clusters displayed on a cortical flat-map. (E) Average cluster profiles. Error bar indicates standard deviation across parcels within each cluster. (F) Cytoarchitectonic structural type co-vary with ΔR2*.
Figure 6 with 3 supplements
The anatomical underpinnings of the vascular network architecture.
(A) Neuron (Collins et al., 2010), (B) total receptor density (Froudist-Walsh et al., 2023), and (C, D) R2* and ΔR2* (current study). Multiple linear regression model was used to investigate the relationship between neuron and total receptor densities and (E) baseline R2* and (F) ΔR2* across layers. T-values are threshold at significance level (p < 0.05, Bonferroni corrected).
Figure 6—figure supplement 1
Comparison with neuron density and baseline R2* and ΔR2* in the entire cerebral cortex.
(A) Neuron density map. Data was obtained from literature (Collins et al., 2010; Froudist-Walsh et al., 2023). (B) Baseline R2* displayed in a representative equivolumetric layer 2a (EL2a) and (C) ferumoxytol-induced ΔR2*, an indirect proxy measure of cerebral blood volume (CBV), displayed in EL2b. Scatter plots of (D) R2* and (F) ΔR2* plotted with respect to the neuron density in EL2a and EL2b, respectively, with red lines indicating linear fits. Pearson’s correlation coefficient between neuron density and (E) R2* and (G) ΔR2* across equivolumetric layers (ELs), with dashed lines indicating 95% confidence intervals. Notably, the R2* intercept deviate from the free water R2* (≈1 1/s), and the ΔR2* intercept is non-zero, suggesting that both R2* and ΔR2* are substantially influenced by non-neuronal factors.
Figure 6—figure supplement 2
Anatomical underpinnings of heterogeneous vascular density.
(A) Dendritic tree size and (B) number of dendritic spines per layer 3 pyramidal cell (Elston, 2007; Froudist-Walsh et al., 2023) are compared with (C) baseline R2* and (D) ΔR2* in equivolumetric layer 3 (3a + 3b). (E) Dendritic tree size and (F) spine counts are negatively correlated with cortical variation in R2* and ΔR2*. The error bars indicate 95% confidence intervals.
Figure 6—figure supplement 3
Cerebrovascular volume varies along the cortical hierarchy.
(A) Level of area within cortical hierarchy. Panel A has been adapted from Figure 3E from Markov et al., 2013. (B) Selected cortical areas displayed on a cortical surface map. (C) Ferumoxytol-induced change in transverse relaxation rate (ΔR2*), an indirect proxy measure of vascular volume, plotted with respect to the hierarchical level. The upper panel displays visual dorsal stream, the middle panel visual ventral stream, and the bottom panel somatosensory. Hierarchical orders were obtained from Felleman and Van Essen, 1991; Markov et al., 2014a.
Author response image 1
(A) T2w-FLAIR SPACE normalized signal-intensity plotted vs neuron density. Note that low signal-intensity corresponds to high R2 and high neuron density, consistent with findings using ME-GRE. (B) Correlation between T2w-FLAIR SPACE and neuron density across equivolumetric layers. Notably, a similar relationship with neuron density was observed using a variable spin-echo pulse sequence as with quantitative gradient-echo-based imaging.
Tables
Table 1
Estimated transverse relaxation rate (R2*) before and after injection of ferumoxytol contrast agent.
Values are mean (std) (N = 4). Abbreviation: WM: white matter.
Additional files
Download links
A two-part list of links to download the article, or parts of the article, in various formats.
Downloads (link to download the article as PDF)
Open citations (links to open the citations from this article in various online reference manager services)
Cite this article (links to download the citations from this article in formats compatible with various reference manager tools)
Mapping vascular network architecture in primate brain using ferumoxytol-weighted laminar MRI
eLife 13:RP99940.
https://doi.org/10.7554/eLife.99940.4
